UNIVERSITY OF WARWICK
SCHOOL OF ENGINEERING
Warwick Mobile
Robotics: Urban Search
and Rescue Robot
ES410 Group Project Technical Report
02/05/2013
Authors:
Kristian Buckstone
Lewis Judd
Nicholas Orlowski
Michael Tayler-Grint
Rachele Williams
Edgars Zauls
Abstract
Urban Search and Rescue (USAR) robots are designed to locate survivors in hazardous
environments such as earthquake zones. The Warwick Mobile Robotics (WMR) team has
enhanced a prototype USAR robot with the aims of:

Entering the World RoboCup Rescue competition

Improving the prototype’s commercial viability by eliminating key weaknesses in the
current design
WMR 2012/2013 has been working to address key weaknesses in the tele-operated robot
design. The system’s reliability has been improved by redesigning the chassis and head to
reduce vibrations and eliminating recurring problems in the electronics. Additionally, system
safety has been increased by developing an easy-to-use battery-pack.
Reuse of existing sensors has increased operator awareness and been a cost-effective way to
markedly improve robot tele-operation performance. This was also boosted by a revamped
controller layout and GUI, which allows the operator to fully utilise the newly-developed
inverse kinematic capabilities.
A full-system test in a realistic outdoors scenario has been carried out at Northrop Grumman
Remotec’s facility. On this basis good performance is predicted in the 2013 World RoboCup
Rescue championship in Eindhoven.
i
Table of Contents
1
Introduction ....................................................................................................................... 1
2
Background to the WMR Project ....................................................................................... 1
3
4
5
6
2.1
Motivation for Development of Search and Rescue Robotics .................................. 1
2.2
Commercial Competition .......................................................................................... 2
2.3
World RoboCup Competition 2013 ........................................................................... 4
Product Development Strategy ......................................................................................... 6
3.1
Requirement Analysis ................................................................................................ 6
3.2
SWOT Analysis ........................................................................................................... 8
3.3
WMR 2012/2013 Objectives ..................................................................................... 9
Chassis ............................................................................................................................. 10
4.1
Initial State............................................................................................................... 10
4.2
Specification ............................................................................................................ 11
4.3
Chassis Redesign ...................................................................................................... 11
4.4
Shell Design ............................................................................................................. 22
4.5
Results ..................................................................................................................... 23
Power System .................................................................................................................. 24
5.1
Introduction ............................................................................................................. 24
5.2
Battery Enclosure .................................................................................................... 24
5.3
Battery Monitoring .................................................................................................. 27
5.4
Power Board ............................................................................................................ 29
5.5
Wiring ...................................................................................................................... 31
Sensors and Devices ........................................................................................................ 33
6.1
Initial State............................................................................................................... 33
6.2
Specification ............................................................................................................ 36
6.3
Actions ..................................................................................................................... 36
ii
6.4
7
Results ..................................................................................................................... 43
Control system................................................................................................................. 44
7.1
Initial State............................................................................................................... 44
7.2
Specification ............................................................................................................ 45
7.3
Architecture ............................................................................................................. 45
7.4
Graphical User Interface.......................................................................................... 47
7.5
Collision System and Error Checking ....................................................................... 48
7.6
3D Representation................................................................................................... 50
7.7
Arm Control ............................................................................................................. 53
7.8
Controller Layout ..................................................................................................... 58
8
Testing ............................................................................................................................. 60
8.1
Aims ......................................................................................................................... 60
8.2
Method .................................................................................................................... 60
8.3
Results ..................................................................................................................... 61
8.4
Analysis of Results ................................................................................................... 70
9
Critical Review ................................................................................................................. 72
10
Conclusions .................................................................................................................. 75
11
References ................................................................................................................... 76
Appendices .............................................................................................................................. 82
Appendix A: Chassis supporting data and calculations ....................................................... 82
Appendix B: Power System support materials .................................................................... 96
Appendix C: Sensors & devices supporting data ................................................................. 97
Appendix D: Control system support materials .................................................................. 99
Appendix E: Other support materials ................................................................................ 105
iii
1 Introduction
Urban Search and Rescue (USAR) robots are designed to locate survivors in hazardous
environments such as earthquake zones. This reduces the risk a human rescue team is
exposed to and allows them to plan the rescue more effectively.
The Warwick Mobile Robotics (WMR) project has developed a prototype USAR robot. This
year the team’s aims are:

Enter the World RoboCup Rescue competition, building on previous teams’
successes at the European level.

Improve the commercial viability of the prototype by eliminating key weaknesses in
the current design.
2 Background to the WMR Project
2.1 Motivation for Development of Search and Rescue Robotics
Several recent disasters have highlighted the need for search and rescue robots, such as:

The 2010 Chilean mining accident where 33 men were trapped underground
(Franklin 2010)

The 2011 Fukushima Accident where a tsunami caused destruction to a nuclear
power plant (World Nuclear Association 2013)
The main advantages to using robots in such situations are (Tadokoro 2009):
1. Robots can operate in conditions that humans cannot, locating survivors that would
have died otherwise.
2. Robots can fit through confined spaces, reducing or eliminating the risk of secondary
damage caused by excavating access routes.
3. Using robots can accelerate rescue operations, increasing the chance of finding
survivors.
1
The WMR team recognises the vital contribution of robots to search and rescue operations
and intends to bring the existing design closer to real-world deployment.
2.2 Commercial Competition
USAR robots must be mobile and benefit from having a manipulator and a range of sensors.
Bomb disposal robots have similar requirements. Some commercial robots in both of these
markets are outlined in Table 1.
Table 1. Competitor analysis
Market
Commercial Competitor
Bomb
Disposal
DRDO – Daksh (Press Trust of India 2011)


Battery-powered, teleoperated, wheeled robot (Figure 1).
Main features: manipulator arm, X-ray scanner, water jet
disrupter, shotgun (for breaching locked doors).
Figure 1. DRDO Daksh.
Northrop Grumman Remotec – ANDROS F6A (Northrop Grumman
2013)


Battery-powered, teleoperated, tracked robot (Figure 2).
Main features: multiple video cameras, manipulator arm
with a range of tools and accessories.
Figure 2. Remotec ANDROS F6-A
2
Disaster
Search
Rescue
Tokyo Fire Department – RoboCue (Nosowitz 2011)
and



Battery-powered, teleoperated, tracked robot (Figure 3).
Function: evacuate casualties from danger-zone.
Main features: ultrasonic sensors, infrared cameras, onboard oxygen canister.
Figure 3. RoboCue lifting a victim (Nosowitz 2011)
Satoshi Tadokoro – Snakebot (Diginfonews 2008)



Electrically-powered, teleoperated, cilia-actuated robot
(Figure 4).
Function: deep-penetration fiberscope.
Main features: nylon bristles, light, speaker.
Figure 4. Snakebot
VECNA Solutions – BEAR (VECNA 2013)



Battery-powered, teleoperated, tracked (Figure 5).
Function: hevy lifting over rough terrain.
Main features: hydraulic upper-body, reconfigurable track
“legs”, versatile manipulator arms.
Figure 5. Visualisation of BEAR Medic-Bot
3
It can be seen that most commercial solutions are battery-powered, teleoperated and
tracked much like WMR’s prototype.
2.3 World RoboCup Competition 2013
The RoboCup Rescue competition is an independent event that allows USAR robots to be
compared based on their performance. The competition is based around a simulated
disaster zone through which robots must navigate in order to find victims. Points are
awarded for each victim discovered as well as for other abilities such as mapping and
manipulation.
For the last five years, WMR has attended the European competition in Germany. Since the
robot has performed consistently well at this level the team has decided to attend the World
competition this year.
4
2.3.1 RoboCup Competitor Analysis
The two best teams from the 2012 World competition are listed in Table 2 as examples of the world-class robots that WMR seeks to compete against.
Table 2. Top performing robots in recent World RoboCup competitions.
Types of robots
used
Competitor Team
MRL (Iran)
Key Features of main competitor robot
Three:





Autonomous
Teleoperated
(Figure 6)
Flying
Sensors
System
Cost
Motor outputs are coupled in
pairs for increased power.
6 degree of freedom (DoF)
manipulator arm. (Shahri, et al.
2011)
CO2 sensor, inertial £6000measurement
unit £10000
(IMU), LiDAR, motor
encoders,
sonar,
thermal
camera,
video cameras
4 wheel drive and steering
(Hector Darmstadt 2012)
Vision system for hazmat
symbol and victim identification.
CO2 sensor, IMU, £12000
LiDAR, microphone,
motor
encoders,
sonar,
thermal
camera, X-Box Kinect
(Graber, et al. 2011)
Figure 6. NAJI-VII robot (Shahri et al, 2012)
Hector Darmstadt (Germany)
Three:



Autonomous
(Figure 7)
Flying
Marine


Figure 7. Hector GV robot (Hector Darmstadt, 2012)
It can be seen that both teams have multiple robots. Of the two teams’ robots, WMR’s design is most similar to the NAJI-VII.
5
3 Product Development Strategy
As outlined in the Introduction, the WMR team’s aims were to enter the World RoboCup
Rescue competition and improve the commercial viability of the prototype. The design
process used to achieve these aims is shown in Figure 8.
Requirements to achieve the aims were linked to strengths and weaknesses of the original
prototype and objectives which provided the largest benefit for the effort required were
given priority. Each sub-system had a specification and test-plan written based on the
objectives chosen.
Figure 8. Approach to the development of the WMR 2013 robot.
3.1 Requirement Analysis
Table 3. USAR requirement analysis
Mobility

Navigate complex terrain

Clear a standard door width of 72cm

In the competition piles of rubble are simulated by step
fields, while ramps may represent collapsed wall or ceiling
sections (Figure 9).
6
Figure 9. Step field and 45 degree ramp in a RoboCup Rescue competition.
Manipulation

Manipulate small items (e.g. a water bottle)

Access victims in narrow crevices at a maximum height of
1.6m (Figure 10 (a))

Build structural shoring to support a column (Figure 10 (b))
(a)
(b)
Figure 10. (a) WMR robot accessing a victim box in 2012 German Open. (b) The new
structural shoring task (Pellenz 2013).
Power

Batteries must endure a 25 minute competition run
Victim
identification

Identify victims using:
o Form
o Visual – eye charts, hazmat labels, QR codes
7
o Heat
o Sound – two way audio
o CO2 – to detect breath
Control

The robot must be autonomous, teleoperated by a single
operator, or some combination thereof.
Mapping
Reliability

Intuitive tele-operated controls.

Operator station easy to set up by one person.

Generate map of the site in GeoTIFF format

Mark victim locations

Real-world applications would require the robot to be both
reliable and robust.
3.2 SWOT Analysis
Testing in the lab and during the Imagineering fair (The Imagineering Foundation 2012) was
performed to develop a SWOT analysis of the existing system (Table 4).
Table 4. SWOT analysis of the existing system.
Strengths



Weaknesses

Highly mobile. Received best in
class award in the 2012 European
competition
Strong and stable manipulator
arm.
Wide range of sensors.




Opportunities


Arm uses joint control – takes
practise.
Poor graphical user-interface
(GUI) layout.
Head suffers from vibrations.
Some devices fail to power up
consistently while others are
completely non-functional.
Batteries are very difficult to
connect/disconnect
and
no
charge information is available.
Threats

The LiDAR and IMU devices could
be used to implement mapping
Inverse-kinematics control for the
arm

8
The poor state of the robot’s
wiring make it difficult to
diagnose faults.
The flipper motor brackets bend


Remotec’s expertise and testing
facilities could be used to
improve the robot’s ruggedness.
Competing
in
the
world
competition may help to attract
sponsors.

during normal use, suggesting a
design flaw.
The software is undocumented
and contains a lot of unused
code, so making changes is timeconsuming.
It was found that the existing system had a large number of useful features in theory.
Despite this the robot’s actual performance was poor, mainly due to low reliability and poor
operator feedback. A number of expensive sensors had been purchased in previous years
but some were not used at all, while others failed to live up to their potential.
3.3 WMR 2012/2013 Objectives
With the above in mind the objectives for the 2012/2013 team were set as follows.
Improve reliability of the system by:

Increasing the strength of the chassis to withstand 0.5 m drops without permanent
damage

Reducing vibrations in the head

Eliminating recurring failures in the electrical system
Improve operator awareness and control by:

Using inverse kinematics to control the arm

Designing a more intuitive control interface

Providing feedback to the operator from all sensors
Progress towards real world readiness by:

Carrying out a full system test in a realistic scenario

Taking measures to proof the robot against the elements

Redesigning the battery system for ease of use
9
4 Chassis
4.1 Initial State
The 2012 chassis was fabricated from steel sheet (Figure 11). Additional strength was
provided by two aluminium brackets and a base-plate (Warwick Mobile Robotics 2012).
Figure 11. Old WMR chassis.
During initial testing a number of issues with the existing chassis were identified:

Deformation due to operating loads leading to excessive vibrations and difficulty in
driving the robot.

Flipper motor brackets were damaged. This has been a recurring issue (Figure 12)
due to the uncertainty of impact loads.

Centre of mass of the robot was set high and towards the back requiring the robot
to be driven in reverse up steep inclines.
2011 and 2012
2010
Figure 12. Previous flipper bracket designs deformed.
10
4.2 Specification
1. Increase strength of the chassis to withstand 0.35 m drops without deformation
whilst reducing weight.
2. Redesign flipper brackets to withstand service loads.
3. Reposition internal systems to move the centre of mass down and forward.
4.3 Chassis Redesign
4.3.1 Load Estimation
Initially, accurate estimates of the loads the robot must withstand during operation were
obtained. For physical testing an inertial measurement unit (IMU) (Xsens Technologies B.V.
2013) was attached and the robot was driven from a step height of 0.35m (Figure 13). This
was then simulated in SolidWorks (Figure 14) (detailed results in Appendix A.1).
Figure 13. Time-lapse from high-speed footage of a drop test.
To validate the virtual test acceleration plots were compared and found to be similar with
peaks of 41ms-2 for the physical test and 45ms-2 for the virtual test. This 9.75% error can be
accounted for by SolidWorks’ limitations such as the need to define contact points and the
fact that bodies are treated as perfectly rigid (in fact some energy is absorbed in elastic
deformation).
11
Figure 14. Time-lapse of a SolidWorks drop test.
A number of tests were carried out using the validated model. The results were used to
calculate the forces experienced by flippers during impact (1520 N) and the force
transmitted to the flipper motor bracket (18.2 kN). These values were used as loads for the
new chassis design.
4.3.2 Space Frame Design Evolution
A structural space frame design was adopted for the new chassis. This provides a
straightforward design for absorbing various impact loads. There are 6 key force transfer
locations: two flipper shafts, two flipper motors and two drive motors. The initial design
concept (Figure 15) shows the basic shape of the frame.
Figure 15. Sketches of the first chassis design concepts.
Aluminium alloy 6082-T6 was chosen as this provides high strength (310 MPa yield (alu
Select 2013)) combined with low weight. 5 mm plate was used with cut-outs in non-critical
areas to further reduce weight (Figure 16).
12
Figure 16. CAD drawing of an early design.
Figure 17. Preliminary frame FEA showing deformation under offset impact load.
By loading the space frame in various scenarios (Figure 17) (see Appendix A.2 for details) the
design was refined by adding/removing material depending on stresses at particular points
(Figure 18). The rounded edges at the front and back were added to provide a sliding guide
to help negotiating steps. The thick band in the centre shares the forces generated by
flippers during impact.
13
Figure 18. Final chassis side design.
4.3.3
Flipper Motor Brackets
4.3.3.1 Bending Force Estimation
Before redesigning the flipper motor brackets the forces involved in bending last year’s
designs were investigated (Figure 19).
Figure 19. Deformation of 2012 flipper motor bracket.
The loading cases for the front and rear flipper motor brackets were different. The rear
motor was mounted above the flipper shaft (Figure 20 (a)). This led to a reduced effective
second moment of area of the bracket causing the stresses to exceed the yield stress of the
aluminium alloy used (see Appendix A.3 for analysis). It was found that the forces to induce
permanent deformation were 15.1 kN for the front bracket and 1.3 kN for the rear bracket.
Moving the rear motor to a lower position (Figure 20 (b)) improved the loading condition
and also brought the centre of mass forward and down.
14
Figure 20. Position of the flipper motor in (a) 2012 design (above the flipper shaft) and (b) 2013 design (next to
the flipper shaft).
4.3.3.2 Flipper Motor Bracket Design
Two major changes were implemented in the new flipper motor bracket design:

Material was upgraded to aluminium alloy 6082-T6 (heat treated and artificially
aged to increase yield strength).

The outline of the part was redesigned to be much simpler and free of stress
concentrations (Figure 21).
Figure 21. Redesigned flipper motor bracket.
A safety factor of 2 was applied in the design process. This means that the new parts can
withstand loads up to 31.6 kN (see calculations below) without deformation:
15
[
(
(
))]
4.3.3.3 Motor Shaft Support Bearing
It was recognized that providing a stronger motor mounting bracket could result in the
impact loads bending the motor shaft. It was essentially unsupported (i.e. a cantilever). This
was addressed by:

Designing a silver steel shaft extension bush (Figure 22)

Supporting it in a bearing capable of supporting flipper loads (up to 4.62 kN) (part
number 61903-2rs1 (SKF 2013))
Figure 22. Motor shaft supporting bearing arrangement (view from the bottom).
16
4.3.4
Other Structural Components
Figure 23. Space frame with additional components.
To complete the structure of the space frame (Figure 23) two supporting base plates were
added. These were evolved from past designs (Figure 24):

Material grade increased to 6082-T6

Excess material was removed to save weight
2011 base plate
2013 base plate
Figure 24. Comparison of initial and redesigned base plates.
To increase the lateral stiffness at the rear of the chassis a bracket was added (Figure 25).
This ties together the drive motors and acts as cross-bracing.
17
Figure 25. Drive motor bracket providing a cross-bracing effect at the back of the chassis.
4.3.5 Flipper Shafts
The flipper shafts had to be lengthened as the chassis had been widened. A split shaft design
was adopted (Figure 26) for ease of assembly and maintenance. Location was provided by a
set of grub screws resting on a flat on the central part of the shaft.
Figure 26. New flipper shaft design.
4.3.6 Internal Layout
The aim of rearranging the internal layout of the robot was to bring the centre of mass
forward and down while maintaining access to critical components. Due to the fact that CAD
data was not available for off-the-shelf components a cardboard model was constructed
(Figure 27) to investigate various layout options.
18
Figure 27. Cardboard model used to investigate internal layout options.
In order to minimize the weight of the head and increase the weather-proofing of the robot
a number of components were moved into the chassis. The width of the chassis was
increased by 24 mm to accommodate the components (Figure 28) with enough wiring space.
Figure 28. Final internal layout.
4.3.6.1 Thermal Modelling
It was intended to provide no vents in the shell covering the chassis in order to prevent the
ingress of dust and water. However, electronic components still required cooling. This can
be done by circulating air within the chassis to move heat from the hot components to the
19
large thermal mass of the chassis. In order to determine the optimum airflow for cooling and
the number and placement of fans SolidWorks flow simulation was used (see Appendix A.4
for details).
A worst case scenario was assumed where drive motors are at full power. It was found
(Figure 29 and Figure 30) that sufficient cooling is achieved by having two small fans blowing
air over motor controller heatsinks and a third large fan forcing air movement throughout
the rest of the chassis (Figure 31). This ensures an even temperature distribution in the
steady-state without any of the components overheating.
Figure 29. Temperature distribution in the robot (one large and two small fans). Side view.
Figure 30. Temperature distribution in the robot (one large fan and two small fans). Top view.
20
Figure 31. Final arrangement of fans.
21
4.4 Shell Design
To cover the newly developed chassis a new non-structural shell was designed. It functions
as a barrier to dirt and moisture and removal provides access to components inside the
robot. It does not have to carry any significant loads as these are supported by the chassis.
The material chosen was ABS (see Appendix A.5).
4.4.1 Panel Design
The stresses induced shell panels were modelled using beam bending theory (see Appendix
A.6 for details). Thickness of 3 mm was chosen as a result (safety factor of 1.97).
The actual panel geometry is a trade-off between the protection of internal components,
accessibility and ease of manufacture (Figure 32 and Figure 33).
Figure 32. Final shell design with cut-outs.
The seams at the bottom of the shell (Figure 33) can be sealed with a silicon compound if the
robot is to be operated outdoors.
22
Figure 33. Final shell design with cut-outs (bottom view).
4.5 Results

Chassis strength increased eliminating deformation during service

Weight decreased by 23%. This extends battery life

Centre of gravity moved forward and down improving slope climbing capability

No permanent deformation of redesigned flipper brackets observed during testing

Better protection from the elements by eliminating air intakes and locating batteries
internally
23
5 Power System
5.1 Introduction
The power system comprises batteries, regulators, switches, fuses, connectors and wiring.
The focus was on improving:

Battery enclosure, where the batteries are located within the robot and connected
to the power system.

Battery management system, which ensures that the batteries’ safe operating limits
are not exceeded and provides information about the batteries to the CPU.

Power distribution board, which regulates the battery voltage and supplies power to
the electronics.

Internal wiring, carrying power and data throughout the robot.
5.2 Battery Enclosure
5.2.1 Issues Identified
1. While attending the Imagineering fair (Ricoh arena, 20th-21st Oct 2012) it was noted
that the robot’s flippers could obstruct access to the battery enclosures. If
emergency stopped, it would be very time-consuming to remove the batteries,
posing a safety hazard.
2. The enclosure had several sharp edges (a safety hazard) (Figure 34).
3. The batteries’ power connectors were awkwardly positioned. As a result, replacing
the batteries could take several minutes.
4. The modular nature of the connector meant that the battery could be connected
with the wrong polarity.
24
Figure 34: Photo of the original battery enclosure
5.2.2 Specification
1. Batteries must be located where they are easily accessible at all times.
2. The enclosure must be free of sharp edges which would pose a risk to wires or the
operator during connection and disconnection of the battery.
3. Replacing the batteries must take less than a minute.
4. It must be impossible to connect the batteries with reverse-polarity.
5.2.3 Actions
Due to the internal layout of the robot it is back heavy. Therefore, locating the batteries
under the front lid is advantageous as it moves the centre of mass forward. Furthermore,
the front of the robot can be opened easily to allow access to power board connections,
computer ports (for debugging purposes) and on/off buttons.
The battery connector should be asymmetric (i.e. connectable only the correct way) and
rated above 24V and 60A. Additionally, data pins are required to transmit data from the
battery management system to the power board over I2C bus. APP SBS75X (Figure 35)
(Anderson Power Products 2012) was chosen as it has been proven as reliable and fits the
specification.
25
Figure 35. APP SBS75X connector rated at 80V, 110A. Dimensions in [] in mm (Anderson Power Products 2012).
The mating/unmating force of the connector is 70N. This was used as a design load to
optimize the form and size of the battery case and holder. The former was fabricated from
sheet aluminium while the latter was 3D printed in ABS (Figure 36).
Figure 36. Redesigned battery-pack and holder.
The manufactured battery cases (Figure 37) are sturdy and easy to insert and remove from
the robot. This is especially important in case of emergencies where a short-circuited or
damaged battery must be removed from the robot quickly. Guiding rails were also added to
the 3D printed part to ensure the connectors are always aligned accurately.
26
Figure 37. Redesigned battery-pack.
5.2.4 Results
1. The batteries are now easily accessible and cannot be obstructed by the flippers.
2. The battery is protected from sharp edges within the housing, and there are no
sharp edges around the handle.
3. Replacing a single battery takes between 10 and 20 seconds.
4. Batteries cannot be connected with reverse-polarity
5.3 Battery Monitoring
5.3.1 Issues identified
1. The only fault-condition monitored was cell under-voltage. Short-circuits and other
faults were not protected against.
2. In the event of a fault the battery had to be manually disconnected to prevent the
battery being discharged further (with potentially catastrophic consequences).
3. No fuel-gauging was implemented.
27
Figure 38: Photo of the original battery monitor
5.3.2 Specification
1. The battery monitor must protect the batteries against cell undervoltage and shortcircuit/overcurrent conditions.
2. In the event of an undervoltage or overcurrent condition, the battery monitor must
automatically disconnect the battery.
3. The battery monitor must supply state-of-charge information to the robot’s control
system.
5.3.3

Actions
A battery monitoring board was designed (Figure 39) using two ICs – one for
protection, the other for fuel-gauging.

The cell-voltage connectors were chosen to be compatible with the existing battery
and charger connectors.
Figure 39: Photo of the new battery monitor
28
5.3.4 Results
1. The BQ77910A (Texas Instruments 2012) provides protection against under-voltage,
excessive discharge current and short-circuit conditions. It can also protect against
over-voltage and excessive charge current but this would interfere with the
regenerative-braking provided by the AX3500 motor controllers (Roboteq 2013) and
so was disabled.
2. Should a fault occur, the chip automatically switches an attached NFET off to
prevent the battery discharging further. Recovery is automatic once the fault
condition is removed.
3. The BQ34Z100 (Texas Instruments 2013) provides fuel-gauging reported to be
accurate to less than 1% of capacity.
5.4 Power Board
5.4.1 Issues Identified
1. The 30W DC/DC converters used on the power board were not capable of powering
all systems at full load. This would cause devices to stop functioning as their power
supply cut out.
2. The output connectors (Figure 40) were not arranged in any obvious pattern.
Adjacent connectors had different voltages and polarities, which was confusing
when connecting devices.
29
Figure 40: Photo of the original power board
5.4.2 Specification
1. Each power supply must be able to power all devices of the supply’s voltage at full
load.
2. Connectors must be grouped according to voltage, with consistent polarity. Both
voltage and polarity must be marked on the board.
5.4.3

Actions
50W DC/DC converters with heat-sinks were used to improve reliability and
operating temperature range.

Small footprint power-FETs reduced the size of the power switching section of the
board.

An mbed microcontroller (mbed 2013) was used to control the board since the
supporting toolset made programming very straightforward.

The pluggable screw-terminals were replaced with Harwin 101-Lok power
connectors for higher reliability and current capacity.

A 4-port I2C interface was added to allow the power board to communicate with the
battery fuel-gauge chips.
30
Figure 41: Photo of the new power board
5.4.4 Results
1. The new power supplies can source the maximum rated current for all attached
devices with 20% spare capacity.
2. Connectors are grouped by voltage, with polarity labelled. Connectors cannot be
connected with reverse-polarity.
5.5 Wiring
5.5.1 Issues Identified
1. There was no wiring diagram for the robot, which made maintenance more time
consuming as connections had to be written down as they were disconnected to
ensure that the device could be reconnected later.
2. The drive and flipper motors’ power supplies were arranged in a large loop around
the base of the chassis, effectively creating a loop antenna (Figure 42). This was a
major concern as it could potentially interfere with not only the robot’s electronics
but also any nearby electronics.
31
Figure 42: Original motor power supply wiring
5.5.2 Specification
1. A wiring diagram must be compiled, detailing all electronic systems within the robot,
their ports and all wires, cables and connectors used to connect them.
2. Systems should be wired so that the current loop produced by their power supply
(traced from either battery’s positive terminal to its negative terminal) is as small as
possible.
5.5.3

Actions
A bus-bar was used to simplify power connections.

The motor controllers were repositioned and rewired to eliminate current loops.

All connections were noted and used to construct a wiring diagram for future
reference (see Appendix B.1).
5.5.4 Results
1. A wiring diagram now exists detailing all connections between systems, including
connector type, pinout and wire colours.
2. Current loops have been minimised.
32
6 Sensors and Devices
For the robot to perform its search and rescue tasks, various sensors and devices are
mounted on the chassis and head. This section outlines the initial state of the robots
capabilities followed by the team’s developments to improve the robot’s functionality.
6.1 Initial State
Figure 43. WMR 2012 robot head.
Figure 44. Existing position measurement sensors (a) on flipper motors, (b) on arm joints.
Sensors and devices were mounted both on the head of the robot (Figure 43) and within the
body (Figure 44).
33
After initial testing, the capabilities of the robot were reviewed in terms of the
devices/sensors used. Table 16 in Appendix C.1 shows the expected capabilities of the robot
with the current devices/sensors that provide this capability and the strengths/weaknesses
of this device in achieving this capability.
From the preliminary analysis, it was clear that in order to increase operator awareness an
accurate portrayal of the flipper positions was necessary. The current encoders were relative
and so could not be used to store the current flipper positions.
There were several sensors and devices that were attached to the robot but were either not
working or had no function. These include:

Gripper module

Infrared camera

IMU

LiDAR
Initial testing showed that each of these devices were either unpowered or had no method
of communication with the client; they were all however in working condition when tested
separately from the robot.
34
6.1.1 Initial Head State
A SWOT analysis was performed on the previous head design (Table 5).
Table 5. Head design SWOT analysis.
The initial head design using sheet aluminium and a rapid prototyped cover had to be
improved due to its weight and chaotic internal layout. The power distribution within the
head was haphazard, illogical and confusing. This could explain some of the intermittent
dropout problems that were occurring. A complete overhaul of the wiring needed to be
completed.
35
The LiDAR and IMU were both previously in the head (increasing weight) yet non-functional,
therefore a decision process was needed to decide whether or not these components should
be kept in the head, moved elsewhere or removed completely (Table 6).
Table 6. LiDAR and IMU placement decision table
The result of this process was the decision to remove the LiDAR and IMU from the head and
place them on/ into the body. The other sensors were all kept in the head, but some
electronic components were moved into the body to once again save weight.
6.2 Specification
To improve the robot beyond its current capabilities the following aims were set:

Make all sensors work and communicate correctly with the GUI.

Improve drivability and operator awareness through better use of the LiDAR, IMU,
flipper encoders and potentiometers.

Redesign of head unit to reduce weight and improve sensor layout.

Reorganise wiring and layout to ensure all sensors and devices work correctly.
6.3 Actions
6.3.1 Implementation of Existing Devices
The LiDAR has been moved from the top of the head to the front of the body. It can be used
with a SLAM algorithm to create a 2-D map of the surrounding environment; giving the
36
operator a clear view of their surroundings. This year, the LiDAR was predominantly used to
provide information on what is directly in front of the robot to aid short-range navigation.
The IMU has been moved inside the body of the robot for orientation data, centre of mass
calculations and stabilization of the LiDAR 2-D map. The calculation of the centre of mass, as
well as diagrammatic display of the robot orientation aid the operator to recognize and take
action if the robot is about to ‘topple’ over.
The gripper has been mounted below the head and supplied with PWM control. An mbed
microcontroller was introduced to the head to supply the robot with the required PWM
control.
The infrared camera has been implemented, providing data to the robot and the power
issues have been solved by rewiring the head.
6.3.2
Sensor/ Device Changes
6.3.2.1 CO2 Sensor
The CO2 sensor circuit board was redesigned to ensure correct driving voltage is supplied.
The circuit design is shown in Figure 45 (board layout in Figure 46). An operational amplifier
is used to amplify the small change in voltage output from the sensor to be sent to the mbed
microcontroller. A voltage regulator steps down the incoming 15 V to 6 V for the sensor, the
capacitor values are calculated from the data sheet. The capacitor C3 is a decoupling
capacitor.
37
Operational Amplifier
C3
Cardon Dioxide Sensor
U2
1
2
Voltage Regulator
100nF
8
U1
1
2
3
3
MC7806C
C1
0.33F
3
6
5
4
3 pin connector
J1
In/Outputs:
Sensor output
GND
15V
U3A
1
2
MG811
LM358N
HDR1X3
R14
C2
0.1µF
5.6kΩ
R2
3.1kΩ
R3
310kΩ
Figure 45. Circuit for CO2 sensor.
Figure 46. Ultiboard layout for CO2 sensor board.
6.3.2.2 Flipper Encoders
Design of a system of absolute encoders to monitor the position of the robot’s flippers was
carried out by previous year’s team (Warwick Mobile Robotics 2012). As the correct parts
were already identified, it was decided to implement this system. For the flipper encoders to
correctly operate two Melexis 90316KDC Hall-effect encoders were mounted on circuit
boards and small magnets were attached to flipper motor shafts (Figure 47). The boards
were connected to the flipper motor controllers to provide absolute position feedback.
38
Figure 47. Encoder circuit mounted on flipper motor bearing housing.
6.3.2.3 mbed Microcontroller
For the CO2 sensor and gripper module to communicate to the central computer an ARM
mbed microcontroller (mbed 2013) was introduced into the head. This is required to:

convert the analogue CO2 senor data to digital data and send it to the computer

receive gripper target position from the computer and supply the correct PWM
control signal to the gripper
Figure 48 shows the circuit of the mbed (board layout in Figure 49). The two resistors are to
drop the high voltage coming in from the CO2 sensor to a voltage suitable to input to the
mbed.
Connections:
J2
5V
U1
GND
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
HDR1X2
J4
GND
Ang IN from
CO2 Sensor
HDR1X2
R3
47kΩ
R4
12kΩ
40
39
38
37
36
35
34
33
32
31
30
29
28
27
26
25
24
23
22
21
J3
mbed
HDR1X3
Figure 48. Circuit layout from mbed board.
39
Outputs to gripper:
PWM
5V
GND
Figure 49. Ultiboard layout for the mbed board.
6.3.3
Head Redesign
6.3.3.1 Design Specification for the New Head

Reduced weight

Reduced size

Must fit and protect all components within

Organised layout, both for components and wires

All components must be easy to access and be able to be removed individually,
without dismantling the head (for quick repairs)

Impact resistant and strong (in comparison to previous design)
6.3.3.2 Easy Access
With 3D printing expertise in the group, it was decided that this would be an appropriate
technology to use as it allows the creation of complex geometry to fit each component
(Figure 50). ABS plastic was used due to its impact resistance and because it’s a standard
material for the 3D printing process used (FDM).
40
Figure 50. CAD design and finished head assembly.
A removable lid was designed to aid with access to components for maintenance. Using a
sliding design rather than a hinge makes the head less susceptible to dust and rainwater
(Figure 51).
Slide
Figure 51. Sliding lid for easy access to internal components.
41
6.3.3.3 Weight Reduction
A breakdown of the weight of the original head (excluding cables) is shown in Table 7.
Table 7. Component contributions to initial head weight.
A breakdown of the weight (excluding cables) of the new head design is shown in Table 8.
Table 8. Component contributions to redesigned head weight.
There is a total mass saving of 29% in the new design which reduces vibrations and makes
the head easier for the arm to lift.
6.3.3.4 Size
The head is taller than the previous design (if the LiDAR is excluded) however this is because
the IP camera did not fit into the old design. The new design is more importantly thinner and
shallower (Figure 52) allowing the head can to achieve 180 degrees of movement in both the
tilt and pan without colliding with the arm.
42
Figure 52. Side to side comparison of 2013 (left) and 2012 (right) WMR robot heads.
6.4 Results

Head weight reduced by 29% thus minimizing vibrations

All sensors now operational:

o
LiDAR and CO2 sensor data displayed in GUI
o
IMU and flipper encoders used for 3D representation
Gripper is functional and can be controlled by the operator
43
7 Control system
7.1 Initial State
From the preliminary testing of the robot, as well as advice from the previous year’s team, a
SWOT analysis of the current control system was performed (Table 9).
7.1.1
SWOT Analysis of Control System
Table 9. SWOT analysis of the initial control system
44
Reliability was highlighted as problem for the previous team, with motors dropping out and
communication problems during preliminary testing. Also, the system was not providing the
driver with information about surroundings which on previous occasions had caused
catastrophic failures (the robot toppling in the 2011 RoboCup competition). It was evident
the robot lacked functionality which could be improved through 3D representation and
heartbeat to immobilize robot if communication lost.
7.2 Specification

Provide operator with awareness of the robot’s surroundings

Improve the reliability of the system by preventing motor drop-outs and
communications failures

Improve the intuitiveness of control by remapping the controller layout and
improving the arm control

Make the system easy to learn by automating and simplifying user controls

Give more information to the driver regarding errors and robot operations
7.3 Architecture
The overall architecture of the software is inherited from previous years (Figure 53).
45
Figure 53. Architecture of the robot control software.
46
7.4 Graphical User Interface
The GUI is the only method to provide the operator with information about the robot. The
SWOT analysis highlighted some weaknesses of the operator interface, the most prominent
being the missing features and lack of organisation (Figure 54).
Figure 54. Initial state of the GUI.
The new GUI holds all of the features and resizes with any display (Figure 55). The layout has
been improved so that panels are fixed in place and cannot ‘float’ around the screen.
Figure 55. Redesigned GUI layout incorporating new types of feedback.
47
7.5 Collision System and Error Checking
It was discovered during testing that despite the joints having limits set so that the arm
should not crash into itself, there was no function to prevent the arm from colliding with
other parts of the robot (Figure 56).
Figure 56. Robot arm crashing into the body using 2012 attempt at inverse kinematics lacking collision
detection.
Preventing simple driver errors like surpassing joint limits and colliding with other parts of
the robot is crucial to the reliability and ease of use of the system.
7.5.1 Collision Detection
One method of detecting collisions is to calculate positions of each free-body in 3D space
and detect whether any intersect. This method is thorough but complex, time demanding
and processor intensive. As collision detection was not a critical objective, a quick method to
detect collisions was found in Java3D library (Oracle 2013).
48
Figure 57. Visualization of the collision boxes used in the collision detection system.
Using this system, a scale model of the robot (with safety factors added) was constructed to
run on the server (Figure 57). This detects if a body part is about to collide with another
before the command is sent to the motor. If so, the function returns a Boolean error variable
and prevents the robot from moving to this position and notifies the operator (Figure 58).
Figure 58. Collision detection algorithm.
49
7.6 3D Representation
The 3D representation provides the driver with awareness of the robot configuration in
space.
Figure 59. 3D representation designed by the 2010 WMR team (not in use last year) (Warwick Mobile Robotics
2010).
An existing 3D representation (Figure 59) was not functional. A SWOT analysis was
performed on the previous design, considering the code itself and its suitability to be
updated.
7.6.1
SWOT Analysis of 3D Representation
Table 10. SWOT analysis of the existing 3D representation
50
The 3D representation was seen as an essential tool for giving the user awareness of the
robot and errors that occur when operating remotely. The SWOT analysis highlighted many
potential areas of improvement including:

Adding flippers to the representation

Visually showing errors

Providing the user with the orientation of the robot – where the ground is and the
angle of the robot.
The threats were considered and alleviated; with one team member being responsible for
learning Java3D.
7.6.2
Updates to 3D Representation
7.6.2.1 Graphics
There were many redundant features on the old graphic, such as a sky and arena. These
elements have been removed and the design has been streamlined for simplicity of
understanding and better contrast in outdoor environments (Figure 60).
Figure 60. Updated 3D representation graphics
7.6.2.2 Flipper Positions
Previously there was no method other than camera positioning to give the user the flipper
positions. Flipper encoder data in now used and the flippers have been added to the 3D
51
representation, much in the same way as the arm, where positions are stored on the server
and then sent across to the client and displayed on the graphic. Having the flipper encoder
data also allows the flippers to be included in the collision detection class.
7.6.2.3 Robot Orientation
The orientation data from the IMU has been added to the 3D representation. A grid was
used to represent a level ground surface about which the robot rotates (Figure 61). This
should help give advance warning of toppling.
Figure 61. 3D representation showing robot angle with reference to the horizontal plane.
7.6.2.4 Collision
The 3D representation was decided to be the best way to show the user the details of any
collisions or joint limits. This was achieved by building in a function which turns the body
parts which will collide from green to red (Figure 62).
52
Figure 62. Potential collision highlighted just before occurring.
7.7 Arm Control
Joint-control was the only method to control the arm on the previous design. Initial testing
studied the ability of the arm to reach certain positions and the ease at which an
inexperienced user could operate the controls.
7.7.1 SWOT Analysis of Arm Control
From initial testing, a SWOT analysis was performed (Table 11).
53
Table 11. SWOT analysis of existing arm control options
It was decided to implement inverse kinematics enabling the user to control the position of
the end-effector within three-dimensional co-ordinate space, rather than joint control. With
this method of controlling the arm the user only has to aim towards the target and the
inverse kinematics calculates the joint angles necessary to get there.
The controller layout was also highlighted as an area of improvement to improve the ease of
use and speed of learning for new users.
7.7.2
Specification for the New Arm Control System

Intuitive controller layout: existing functionality as well as new features

Implement inverse kinematics to allow the user to ‘fly the head’

Keep track of arm position at all times to allow transition between control methods
and provide feedback of arm position to 3D representation

Smooth motion keeping overshoot to a minimum
7.7.3 Inverse Kinematics
Previous attempts (Warwick Mobile Robotics 2012) used matrix methods to solve the IK
equations however this year the two planes of motion have been taken separately and then
combined with the equations, making the problem simpler.
54
7.7.3.1 Cylindrical Co-ordinate Frame
The kinematics was first solved in a cylindrical co-ordinate frame and the transformed into a
Cartesian co-ordinate frame. A cylindrical co-ordinate frame was chosen because the arm
effectively acts in a two-dimensional plane r-z which is rotated about the z axis by the turret
(global angle ).
Consider the arm moving in just the r-z plane with end-effector co-ordinates
and
(Figure 63).
r
Figure 63. Arm in the Cartesian coordinate frame.
Using Pythagoras, the distance
target positions
and
from the origin and angle
:
√
( )
Using trigonometry,
,
, and
, are calculated:
55
are calculated in terms of the
(
(
(
)
)
)
These functions were implemented into the software to calculate the angle of the shoulder
and elbow joints (as well as the tilt angle) needed to reach any position in the x-z plane. In a
cylindrical co-ordinate system of r-z- , the position f the robot arm in three dimensions can
now be achieved by setting the turret rotation
to the global angle
(Figure 64).
Figure 64. Top down view of the cylindrical coordinate frame.
7.7.3.2 Cartesian Co-ordinate Frame
In order to fly the head in all directions, it was necessary to transform the cylindrical coordinate system into a Cartesian system. As both systems have the same z co-ordinate, the
only transformation that needs to occur is the two-dimensional transformation from r- to
x-y.
56
Figure 65 Robot arm top-down view (Cartesian Co-ordinate frame)
From trigonometry:
These transformations provide all the information needed to find the joint angles to move
the head to any x-y-z co-ordinate. If x, y and z co-ordinates are provided, the value of r can
be calculated from x and y. The kinematics are then solved in the cylindrical co-ordinate
frame and the arm moves.
7.7.3.3 ‘Flying’ the Head
The operator should be able to move directly at a target object that the robot head is facing.
This was achieved by creating a new local co-ordinate frame based upon the tilt and pan of
the end-effector (see Appendix D.1).
7.7.4 Linear Interpolation
In order to make the movement of the arm smooth and on a straight path, linear
interpolation between the current and target point is used for all movements. This simply
uses Pythagoras’s theorem to calculate a series of points between the target and position
57
which have small intervals between them, therefore making the arm move in a relatively
straight path.
7.7.5 Pre-sets
The previous design had two ‘pre-set’ positions for the arm which could be accessed from
the controller. These allowed the arm to be placed either low-down and protect it during
driving whilst the other raised the arm in order to search for targets. These pre-set positions
were useful and were adapted to work with the new inverse kinematics code (by setting the
pre-sets based on co-ordinate positions rather than joint angles).
An extra pre-set position ‘Limbo position’ (Figure 66) was devised and added to the GUI–
this is both forwards and backwards and allows the arm to be stretched as far as possible
straight out in front of the robot.
Figure 66. Robot in “limbo” position under a low table.
As well as adding pre-sets, a new feature this year was created that allows the user to store
previous positions of the arm, useful for completing repetitive actions.
7.8 Controller Layout
From initial testing and the SWOT analysis, the controller layout was highlighted as a key
improvement area, the unintuitive layout being a major concern.
Four different controller modes were developed (see Appendix D.2 for details):
58

Drive mode – for navigating complex terrain

Inverse kinematics mode – to fully control the arm for gripping/manipulation

Search mode – mimics controller layout in first-person shooter games

Joint mode – for backup control of the arm
59
8 Testing
8.1 Aims

Compare the modified robot design’s performance to the initial design.

Determine whether the 2013 aims and objectives have been achieved, and to what
extent
8.2 Method
A structured testing plan was devised that ensured results obtained were measureable and
comparable. The tests assessed both the performance of any new features the WMR team
had developed this year (as outlined in this report) and also the performance of the robot in
relation to the capabilities required in the RoboCup competition (IEEE Robotics and
Automation Society 2013).
The tests were performed at the testing facilities of WMR sponsor, Remotec (Northrop
Grumman 2013). This allowed for a more environmentally realistic assessment, with more
obstacles than WMR would have had access to at the University.
Two types of tests were conducted:

Line-of-sight tests where capabilities of the robot only were assessed (Figure 67 (a))

Remote (blind) tests where operator control interface was assessed (Figure 67 (b))
Figure 67. Operator setup (a) for line of sight testing, (b) for remote testing.
60
8.3 Results
Table 12. Testing results and observations
Challenge
Train tracks
Test relation to
Competition
Competition
Capability
Definition of Success
Capable?
Real rubble similar to Mobility
The robot is able to traverse the train Yes
competition
tracks
simulated rubble.
Notes: Test successful over train tracks testing area. However, once
completed and out of test section, the robot locked into forward drive
mode and had to be E-stopped.
Operator
Control
The robot is able to traverse the train Yes
tracks when controlled from a remote
location
(a)
Notes: Test successful over train tracks testing area. However, when the
robot reached following terrain, mud and twigs were caught between
the flippers and tracks; causing a jam. Test was regarded a pass due to:
(b)
 Test assessed mobility and robot was able to traverse rail tracks
 Indoor competition, so no mud/twigs would clog the flippers.
Figure 68. The robot ascending (a) and descending (b) a rail
track
61
Uneven pavement
Terrain is uneven Mobility
under
competition
conditions
The robot is able to traverse the uneven No
road without toppling or inner
components becoming dislodged
Notes: Test failed due to the power cutting out whilst the robot was
travelling over the pavement. A diagnostic afterwards proved this fault
to be due to a metal component causing a short circuit on the power
board due to vibration.
Figure 69. The robot traversing the uneven pavement
Tiled pavement
See above:
Alternative
topography to the
uneven pavement
Mobility
The robot is able to traverse the uneven Yes
road safely
Notes: The test failed initially due to significant connection lag.
However, the definition for passing this test was focused on the mobility
of the robot. The test was passed the second time.
Figure 70. The robot traversing the tiled pavement
62
Range Test
Remote driving during Mobility
competition for long
periods
Distance: 20m
The robot is able travel the entire Yes
distance without cutting out due to any
motor, wiring or battery faults
Notes: Test passed with no problems encountered.
Operator
Control
(a)
The robot is able travel the entire Yes
distance when controlled from a remote
location
(b)
Figure 71. Operator controlling the robot (a) in line of sight
(b) blindly.
Notes: As long as antenna had line of sight no problems were
encountered.
63
Stairs (38o incline)
Slopes of up to 45o Mobility
found in competition
Robot is able to ascend and descend the Yes
stairs without toppling
Notes: Both the ascent and descent were passed without issues.
(a)
Operator
Control
Robot is able to ascend and descend the No
stairs when controlled remotely
Notes: The descent was passed. The ascent test was failed due to a pretest fault; locking the arm’s shoulder joint into an upright position This
raised the centre of gravity and caused the robot to topple backwards.
(b)
Figure 72. : (a) Mobility test descent (b) blind ascent with
raised arm
64
Ramp (30o incline)
Slopes of up to 45o Mobility
found in competition
Robot is able to ascend and descend the Yes
ramp without toppling
Notes: Both the ascent and descent were passed without problems.
There was sufficient traction to hold the robot in place.
(a)
Operator
Control
Robot is able to ascend and descend the Yes
ramp with traction, when controlled
remotely
Notes: Both the ascent and descent were passed without problems –
even with the shoulder joint fault.
(b)
Figure 73. (a) Mobility test descent (b) blind descent
65
Curb (300mm)
Similar to the
competition step
fields
Mobility
The robot is able to ascend and descend Yes
the curb
Notes: Both the ascent and descent were passed without problems.
(a)
Operator
Control:
The robot is able to ascend and descend No
the curb when controlled remotely
Notes: The robot toppled on both ascent and descent. This was due to
the shoulder joint locking fault. The centre of gravity information
provided to the operator did not prevent this failure.
(b)
Figure 74. (a) Mobility curb ascent (b) blind curb ascent
66
Vehicle observation
In the competition
narrow holes must be
accessed.
(a)
Manipulation
The robot is able to manoeuvre its Yes
flippers, arms and head to observe into
a vehicle through an open window.
Notes: Due to lack of operator experience, this test was performed
slowly but the objective of accessing the inside of a vehicle was achieved
(b)
Figure 75. (a) Manipulation test stage 1 (b) stage 2
67
Water bottle gripping (750mm table)
(a)
(b)
Gripper capabilities Manipulation
The gripper is able to grip the bottle from Yes
will be tested in the
a height before releasing it on a target.
competition
Notes: Some initial difficulty manipulating the head to the correct
position, the gripper itself had smooth transitions between the set
increments of opening/closing.
Operator
Control
The gripper is able to grip the water No
bottle from a before placing on a target,
whilst being operated remotely.
Notes: This test was performed in the car, from the same height as the
manipulation challenge. The task was failed due to the robot not being
driven close enough to the car (operator inexperience). Therefore when
the arm was fully outstretched to reach the bottle, it could not reach the
target without surpassing joint limits.
Figure 76. (a) Placing water bottle in a target area (b)
gripping bottle from table
68
Figure 77. Blind gripper test from car seat
Small steps
Maximum step height: 200mm
Basic terrain
navigation
preparation for
competition
Mobility
The robot was easily and quickly able to Yes
ascend these obstacles without toppling
Notes: Test was passed quickly and no problems encountered.
Figure 78. Small step mobility test ascent
69
8.4 Analysis of Results
Most mobility, manoeuvrability and manipulation tests were passed. The exceptions to this
were the uneven pavement negotiation, blind curb and stairs ascents and the blind water
bottle test.
The uneven pavement test was least strongly correlated to the competition tasks. Therefore,
its failure was not critical to the team. Although the robot is required to negotiate complex
terrains that simulate disaster zones in the competition, this particular pavement
topography is not directly assessed. However, designing for resistance to vibrations is
strongly advisable so a “flight check” of all connections and components inside the robot
should be carried out before each run. This test was one of the first conducted; so the fact
that the robot continued through testing after undergoing such vibrations shows robustness.
The blind curb and stairs ascent tests were failed due to a pre-testing fault that damaged the
shoulder motor and caused the arm’s shoulder joint to lock into an upright position. This
raised the centre of gravity, made the robot unstable and caused toppling. As one of WMR’s
key aims was to improve the reliability, it was deemed these were two serious fails. In order
to remedy this issue before the competition, the elbow joint will be repaired and the code
modified to include more rigorous joint limits.
The line-of-sight bottle test was successful. As this is a direct requirement of the
competition, the points for the challenge would have been awarded.
The blind bottle test was failed: The operator misunderstood how close the tracks were to
the car. This led to not being able to reach the bottle. Therefore, additional operator practise
is required until this task can be performed reliably.
The testing has provided a good prediction for how the robot will perform at the
competition, as many of the tasks carried out will need to be repeated at the RoboCup. As
extensive structured testing has not been performed by other WMR teams before, this not
70
only prepares this year’s team for the competition but also allows for a benchmark for next
year’s team to measure progression.
Whilst at Remotec feedback was obtained from their Programme and Future Development
Manager, Peter Green. He was impressed and said that “the WMR system performed well
for its first time out; the railway lines were negotiated with little effort followed by some offroading and then stairs all within minutes of arrival. As the day went on driver confidence
grew and negotiation of terrain and use of arm became second nature.” (Appendix E.1 full
quote).
71
9 Critical Review
The chassis redesign has significantly reduced the weight of the structural components but
the overall weight of the robot is still around 40kg due to the heavy arm and motors.
Reducing this would increase battery life. The shifted centre of mass allows for easier
climbing of steep inclines, with the arm in the stowed position. With the arm damaged and
immobile (as it was during testing) robot mobility is hampered. Chassis strength has proved
to be sufficient in most directions. An unforeseen sideways loading due to excessive track
tension (caused by ingress of dirt) caused some deformation in the chassis side. This can be
remedied by adding cross-bracing at the back of the chassis. Furthermore, techniques to
keep the tracks clean of rubble should be investigated.
The improved power system has performed reliably. This is expected to have a very positive
impact on competition performance as wiring problems and sensor drop-outs have
traditionally been an issue. Designing and manufacturing a reliable wiring harness has been
one of the most difficult tasks this year. While the current design performs well it is difficult
to modify and certainly difficult to comprehend for someone who has not been involved in
the process. It is suggested that in the future specialist technical advice on wiring harness
design be sought; one solution would be to use SolidWorks Routing plug-in (Dassault
Systemes 2012). The newly developed battery housings enable much easier, safer and faster
replacement of batteries. Unfortunately, the new battery monitoring system was not
functional during testing due to a communications fault. This will be investigated in more
detail before the competition, but as a contingency the old battery monitor can still be
utilized.
Overall the reliability of the robot has been improved. The start-up was fairly consistent,
with all functions working on the majority of occasions. Therefore, the objective “improve
reliability of the system” has been achieved.
72
The cost-effective approach of using existing sensors to full extent rather than purchasing
new ones has paid off. LiDAR raw data has proven to be a useful driving aid to ascertain
distances to obstacles. Similarly, integrating the IMU data in the 3D representation aids the
operator to avoid toppling. The existing gripper has also proved to be useable, although
operator training is crucial to perform gripping tasks reliably. The CO2 gas sensor has been
unreliable, as it tends to produce random voltage spikes and has long settling times. Running
the warning system on a moving average value instead of the immediate could be useful.
Control lag was observed during testing when the robot was behind a metal staircase.
Generally, the current range of the robot is too small for real-world disaster scenarios.
Improved antennae and transmitters should be investigated. One approach would be to
move the router from the arm base to the inside of the robot and have a set of external
antennae.
A key improvement this year has been the revamped control interface. Providing as much
information as possible in a clear and comprehensible way to the operator is crucial to carry
out successful remote operations. This was accomplished by a 3D representation and a
LiDAR data display. Additionally, “search mode” controller layout combined with working
inverse kinematics enables an extremely intuitive control. Nevertheless, operator training is
still crucial. For example, an understanding of how to best use the flippers when overcoming
obstacles can only be acquired through experience. Flipper presets and simple autonomous
behaviours (e.g. “climb step”) could be designed to assist with this. The mobility,
manoeuvrability and manipulation capabilities of the robot have remained fairly consistent
from last year. Overall the objective “improve operator control” can be assessed to have
been a success.
Through performing the testing in a real world scenario, the team was able to assess how
well the robot could cope with the elements. Successful features included an enclosed shell
and internal battery enclosures, which kept out spots of rain during the test. However, the
73
robot was not able to cope with the mud and leaves that got caught in the flipper tracks
after the rail track test. Debris entering the track system will exert some forces on the
chassis that static stress modelling will not show. Therefore, this objective “progress to real
world readiness” has only been partly successful.
The testing shows that the robot is performing well overall when compared to last year’s
design and also against the competition specifications. It can therefore be concluded that
the team has made good use of the sponsorship given this year. The materials bought with
money donated by the sponsors have been carefully planned, and this is reflected by the
high performance demonstrated through testing.
Throughout manufacture and testing, several potential improvements for the robot were
identified (see Appendix E.2 for details).
74
10 Conclusions
WMR 2013/2013 has been working to address key weaknesses in the tele-operated robot
design. The reliability of the system has been improved by redesigning the chassis and head
to reduce vibrations and eliminating recurring problems. Additionally, safety of the system
has been increased by developing an easy-to-use battery-pack.
Reuse of existing sensors to provide increased operator awareness has been a cost effective
way to produce a marked improvement in robot remote operation performance. This was
also boosted by a revamped controller layout and GUI, which allows the operator to fully
exploit the newly-developed inverse kinematic capabilities.
A full system test in a realistic outdoors scenario has been carried out at Northrop Grumman
Remotec facility. This showed good overall performance corresponding to the aims and
objectives of the project and the competition specification. The materials purchased with
money donated by the sponsors have been carefully planned, and this is reflected by the
high performance demonstrated through testing. This serves as a basis for the prediction of
good performance in the 2013 World RoboCup Rescue championship.
75
11 References
alu
Select.
Mechanical
Properties
EN
AW-6082.
2013.
http://aluminium.matter.org.uk/aluselect/09_mech_browse.asp? (accessed April
28, 2013).
Anderson Power Products. SBS Touch Safe 2-3 Pole Connector with Auxiliary. 2012.
http://www.andersonpower.com/products/multipole-sbs.html (accessed December
5, 2012).
Axis
Communications.
Axis
M1054
Network
Camera.
2013.
http://www.axis.com/products/cam_m1054/ (accessed April 21, 2013).
Bay
Plastics
Ltd.
Plastic
Fabrication:
Industrial
&
Commercial.
2013.
http://www.bayplastics.co.uk/fabrication.htm (accessed April 14, 2013).
Dassault
Systemes.
SolidWorks
Routing
Overview.
2012.
http://help.solidworks.com/2012/English/SolidWorks/sldpiping/c_routing_overview
.htm (accessed April 21, 2013).
Diginfonews. “Active Scope Camera For Use In Emergency Situations : DigInfo.” Youtube. 10
June 2008. http://www.youtube.com/watch?v=djreEBKWPSw (accessed April 10,
2013).
EasyCap. Introducing EasyCap. 2013. http://www.easycap.co.uk/ (accessed April 21, 2013).
Elmira. Elmira Bio-Based Technology. 2013. http://www.elmira.co.uk/ (accessed April 14,
2013).
FLIR
Systems
Co
Ltd.
Photon
160.
2012.
http://www.flir.com/uploadedFiles/CVS_APAC/Cores_and_Components/Photon%20
160%20R-V-Asia%20LR.pdf (accessed April 21, 2013).
76
Franklin, Johnathon. Trapped Chilean miners found alive - but rescue will take four months.
23 August 2010. http://www.guardian.co.uk/world/2010/aug/23/trapped-minersalive-chile (accessed April 9, 2013).
Futurlec.
Technical
Information
-
Carbon
Dioxide
Sensor.
2013.
http://www.futurlec.com/CO2_Sensor.shtml (accessed April 21, 2013).
Graber, Thorsten, Stefan Kohlbrecher, Johannes Meyer, Karen Petersen, and Oskar von
Stryk. “RoboCupRescue 2011 - Robot League Team Hector Darmstadt (Germany).”
SIM.
2011.
http://www.sim.informatik.tu-
darmstadt.de/publ/download/2011_tdp_hector.pdf (accessed April 17, 2013).
Hector Darmstadt. Hector GV. 2012. http://www.gkmm.tu-darmstadt.de/rescue/?q=node/3
(accessed April 17, 2013).
Hitec.
HS-422
Deluxe
Standard
Servo.
2013.
http://www.hitecrcd.com/products/servos/analog/standard-sport/hs-422.html
(accessed April 21, 2013).
Hokuyo Automatic Co. Scanning range finder (SOKUIKI sensor). 2009. http://www.hokuyoaut.jp/02sensor/07scanner/urg_04lx.html (accessed April 21, 2013).
IEEE Robotics and Automation Society. “RoboCupRescue Robot League Rules for 2013.” SSRR
Summer
School.
18
January
2013.
http://wiki.ssrrsummerschool.org/doku.php?id=rrl-rules-2013 (accessed March 16,
2013).
Kolbrecher,
S.,
and
J.
Meyer.
ROS
package
summary:
hector_slam.
2013.
http://www.ros.org/wiki/hector_slam (accessed April 21, 2013).
KS
Composites.
About
Us.
2013.
http://www.kscomposites.com/index.php?option=com_content&view=article&id=1
4&Itemid=60&lang=en (accessed April 14, 2013).
77
Lankhorst Pure Composites b.v. PURE. 2013. http://www.pure-composites.com/ (accessed
April 14, 2013).
Maxon motor ag. 2013. http://www.maxonmotor.com/maxon/view/content/company
(accessed February 14, 2013).
mbed. mbed NXP LPC1768 Getting Started. 2013. https://mbed.org/handbook/mbed-NXPLPC1768-Getting-Started (accessed April 21, 2013).
—. Prototype your product. 2013. http://mbed.org/prototype/ (accessed April 30, 2013).
Northrop Grumman. 2013. http://www.northropgrumman.com/uk/ (accessed February 14,
2013).
—.
Remotec
Robotic
Platforms
and
Sub-Systems.
2013.
http://www.northropgrumman.com/Capabilities/Remotec/Pages/default.aspx
(accessed April 29, 2013).
Nosowitz,
Dan.
Earthquake
Rescue
Robots.
3
March
2011.
http://www.popsci.com/technology/gallery/2011-03/earthquake-rescue-robots
(accessed April 10, 2013).
Oracle.
Java
SE
Desktop
Technologies.
2013.
http://www.oracle.com/technetwork/java/javase/tech/index-jsp-138252.html
(accessed April 2013, 21).
Pellenz,
J.
RoboCupRescue
Robot
League
Rules
for
2013.
2013.
wiki.ssrrsummerschool.org/doku.php?id=rrl-rules-2013 (accessed April 3, 2013).
Phidgets Inc. What are Phidgets? 2013. http://www.phidgets.com/ (accessed April 21,
2013).
78
Press Trust of India.
Army acquires
indigenous
mobile robot 'Daksh'.
2011.
http://ibnlive.in.com/news/army-acquires-indigenous-mobile-robot-daksh/21347711.html (accessed April 29, 2013).
RoboCup Federation. Robot League. 2013. http://wiki.robocup.org/wiki/Robot_League
(accessed April 3, 2013).
RoboCup.
Rescue
Robot
League
Results.
2012.
http://www.robocup2012.org/comp_RCRescue-robot-Results.php (accessed April
17, 2013).
Roboteq. AX3500: 2x60A/1x120A . 2013. http://www.roboteq.com/brushed-dc-motorcontrollers/ax3500-dual-60a-brushed-dc-motor-controller (accessed April 21, 2013).
ROBOTIS. RX-64. 2010. http://support.robotis.com/en/product/dynamixel/rx_series/rx64.htm (accessed April 21, 2013).
S&S.
ABS
Material
Data
Sheet.
2013.
http://www.sselec.com/data/ins%20specs/ABS%20Data%20sheet.pdf
(accessed
April 14, 2013).
School of Engineering. Engineering Data Book. 6th Edition. 2008.
Shahri, A. M, et al. “RoboCupRescue 2011 - Robot League Team MRL Rescue Robot (Iran).”
Vahid
Khorani.
2011.
http://vahidkhorani.com/wp-
content/uploads/2012/11/RRL_MRL_TDP.pdf (accessed April 17, 2013).
SKF.
Deep
groove
ball
bearings,
single
row
.
2013.
http://www.skf.com/group/products/bearings-units-housings/ball-bearings/deepgroove-ball-bearings/single-row/index.html?prodid=1010281903 (accessed April 28,
2013).
Tadokoro, Satoshi. Rescue Robotics : DDT Project on Robots and Systems for Urban Search
and Rescue. London: Springer, 2009.
79
Texas Instruments. 1s to 16s Impedance Track Fuel Gauge | Battery Gas Gauge. 2013.
http://www.ti.com/product/bq34z100?DCMP=analog_power_mr&HQS=bq34z100pr-eu (accessed April 21, 2013).
—.
Standalone
4-10
Cell
Precision
Protector
for
Li-Ion
Chemistries.
2012.
http://www.ti.com/product/bq77910a (accessed April 21, 2013).
The Imagineering Foundation. Fairs. 2012. http://imagineering.org.uk/fairs/information/
(accessed April 2, 2013).
VECNA. Bear. 2013. http://www.vecna.com/innovation/bear (accessed April 10, 2013).
Vishay.
Model
357
product
information.
2009.
http://www.vishay.com/resistors-
variable/list/product-57059/ (accessed April 21, 2013).
Warwick Mobile Robotics. Technical Report. School of Engineering. University of Warwick,
2010.
Warwick Mobile Robotics. Urban Search and Rescue Robot. ES410 Group Project Technical
Report, School of Engineering, University of Warwick, 2012.
World Nuclear Association. Fukushima Accident 2011. 2 April 2013. http://www.worldnuclear.org/info/Safety-and-Security/Safety-of-Plants/Fukushima-Accident2011/#.UW_2yit37l2 (accessed April 9, 2013).
WRevGP.
Electric
Vehicle
Grand
Prix
(evGP).
2013.
http://www2.warwick.ac.uk/fac/sci/eng/meng/warwickracing/evgp/ (accessed April
14, 2013).
Xsens Technologies B.V. MTi. 2013. http://www.xsens.com/en/general/mti (accessed April
21, 2013).
Youngblood, J. Associate Professor of Materials Engineering at the University of Purdue.
Visiting Professor at Warwick. Personal communication. 27 November 2012.
80
https://engineering.purdue.edu/MSE/People/ptProfile?id=11541 (accessed April 14,
2013).
81
Appendices
Appendix A: Chassis supporting data and calculations
A.1 Physical and virtual drop test results
A.1.1 Physical Testing
High-speed footage (Figure 79) was recorded during the physical tests to enable detailed
analysis of impact behaviour. The footage shows the large vibrations on the front flippers
when they hit the floor. Similarly, flexing in certain point in the chassis was observed, which
led to excessive vibrations of the arm and the head.
Figure 79. Time-lapse from high-speed footage of a drop test. Full video available on
http://www.youtube.com/watch?v=okQl-gE5u3M&hd=1
The accelerometer data (Figure 80) shows a maximum peak of acceleration of around 41ms2
, which then reduces in a decaying exponential envelope before becoming steady once
more.
Time (s)
Figure 80. Sample accelerometer data from a drop test.
82
A.1.2 Virtual Testing
A virtual model was set up in SolidWorks (Figure 81) replicating the physical tests carried
out.
Figure 81. Time-lapse of a SolidWorks drop test.
It was found (Figure 82 (a)) that the maximum acceleration closely matches that found in the
physical tests. The maximum force on the right flipper (that which hits the floor first) was
found to be 1520N (Figure 82 (b)).
(b)
(a)
Figure 82. Sample virtual drop test results: (a) Peak acceleration, (b) Peak force
In order to calculate the maximum force encountered by the flipper brackets going through
the main body of the chassis, the following calculation was adopted:
83
It should be noted that the forces calculated by SolidWorks are likely to be higher than in
reality because of the need to model the body rigidly that does not account for some force
being absorbed by the flippers deforming elastically.
84
A.2 Space frame FEA results
The final chassis design is shown below in Figure 83. Compared to the chassis for the 2012
team it was found that the difference in mass was 1kg (23% weight reduction).
The strength of the new space frame sides were tested in SolidWorks Simulation. The model
had to be simplified as the software was unable to compare the whole assembly including
the fixings. Table 13 shows the results of these tests.
Figure 83. Final space frame deformation under side load.
Figure 84. Old chassis side under side load.
Table 13. FEA results comparing the old and the new chassis side
Direction of force
Design
Deflection
(mm)
Percentage of yield
strength of material
Force applied from front
(compressing the frame)
New
0.0012
0.14%
Old
0.011
0.28%
Force applied from above
compressing the frame)
New
0.0026
0.13%
Old
0.089
0.78%
Side force on the front and
New
5.51
12.84%
85
fixed at the back
Old
14.5
189.69%
It can be seen from the table that the new chassis stronger in all directions, especially when
a sideward force is applied. This is the most important direction as the robot was suffering
from the chassis deformation permanently and thus excessive vibrations where caused. By
increasing its strength in this direction with a high yield material and strong frame this
should be avoided, and protect some of the vital electronics which could suffer.
86
A.3 Flipper bracket load estimation
Table 14. Variables used in the calcualtion
Parameter
Definition
Force (F)
Second moment of area (m4)
Distance from neutral axis (m)
Length (m)
Yield stress (Pa)
Thickness (m)
Width (m)
Bending moment (Nm)
A.3.1 Rear bracket
87
A.3.2 Front Bracket
(
(
))
Due to the front flipper bracket being lower in the chassis and the forces being directed
horizontally have to be 15 times more that in the rear bracket.
88
A.3.3 Twisting of the flipper motor shaft
Upon disassembly of the flipper drive system it was found that the keyed motor shafts had
sustained significant damage. Interestingly, the shafts had not worn – they had twisted. By
examining the keyway the angle of twist was estimated to be around 5°. In order to establish
the load that caused this damage the following equation can be used:
where
– torque applied (
2nd moment of area,
),
– resulting angle of twist of the shaft (
– material shear modulus (
),
), – polar
– length of the shaft in torsion ( )
(School of Engineering 2008).
Polar 2nd moment of area for a round shaft can be calculated by:
where
– diameter of the shaft ( ).
So in this case it can be estimated that the 25mm long shaft 12 mm in diameter made out of
carbon steel (
) has experienced a torque of:
It should be noted that this is a conservative estimate of the peak load as the shaft has a
tapered centre hole and a keyway that would reduce its real value.
89
A.4 Thermal modelling results
The following heat sources were modelled:

Drive motors

Power switching transistors on motor controller boards (with heatsinks)

CPU (with small fan) and Northbridge (with heatsink) on computer board

DC-DC power converters on power board (with heatsinks)
Single large fan
-1
Maximum flow through motor controllers: 0 - 0.1 ms
o
o
Motor controllers temperature: 72 C
Drive Motors temperature: 77 C
This first iteration of the fan flow showed that air was being circulated in the front section of
the robot, which is needed to cool the computer and power board. However there is
minimal flow around the motors and the motor controllers, which lead the adding more fans
to see if a better circulation could be generated.
90
Single large fan plus one extra to increase flow through motor controllers.
Maximum flow through motor controllers: 0.35ms
o
-1
o
Motor controllers temperature:64 C
Drive Motors temperature:71 C
The second fan was placed on parallel to the motor controllers, to pass air though the heat
sinks, this increased the flow through the motor controllers but still the model showed no
flow around the motors yet there was a cooling of 6oC which was noticed in the heat plot,
when compared to the previous iteration.
91
Signal large fan plus two to increase flow through motor controllers.
Maximum flow through motor controllers: 0.5ms
o
-1
o
Motor controllers temperature:53 C
Drive Motors temperature:66 C
A third fan was added to the model to see if the flow in the entire robot could be increased.
Ideally this fan would be placed facing the motors to help cooling however due to the lack of
space, this fan was placed on the opposite side to the second fan. This increased the flow
throughout the robot and cooled the motors and motor controllers by 110C each. By using
this configuration the air inside can be cooled on chassis walls using them as heat sinks, this
means the if the model is correct no vents will need to be added to aid the cooling system.
92
A.5 Shell material selection
A number of materials were evaluated for the shell (Table 15). It is advantageous if the shell
is light as this reduces the energy required to move the robot and helps save battery life. It
must also be impact resistant as it will often come into contact with the environment, for
instance, wooden blocks of the step fields. These have edges against which the robot
internals must be protected. An additional requirement is that the shell can be
manufactured easily and any modifications can be carried out in-house. Due to the
uncertainty of using external suppliers for custom components precautions must be taken in
case delivery is delayed or of poor quality.
Table 15. Advantages and disadvantages of the evaluated materials.
Material
Advantages
Disadvantages
Carbon fibre panels
Light
Expensive
Difficult to shape in-house
Tegris polypropylene
Light
composite (Lankhorst Pure Impact resistant
Composites b.v. 2013)
High strength
Impossible to shape inhouse
Bamboo composites
(Elmira 2013)
Strength
insufficient
Very light
Expensive
could
be
Impossible to shape inhouse
Expensive
High Impact Polystyrene
(HIPS)
Impact resistant
Low yield strength
Low melting point (can be
shaped with a heat gun)
Relatively cheap
Acrylonitrile butadiene
styrene (ABS)
Widely used (processing is Scratches easily
well understood)
Relatively cheap
A number of material samples were ordered and tests were carried out (Figure 85). These
included strength tests and bending using a heat gun.
93
Figure 85. High Impact Polystyrene (HIPS) samples bent during testing using a heat gun.
As none of the team members had experience with these materials expertise was sought
with academics (Youngblood 2012), composite component manufacturers (KS Composites
2013) and plastics fabricators (Bay Plastics Ltd. 2013). ABS and HIPS were shortlisted in
favour of the more advanced materials due to the ease with which they can be shaped both
at a supplier and in-house. The manufacturer (Bay Plastics) recommended the use of ABS
over HIPS, so this was chosen as the final material.
A.5 Shell panel load calculations
To determine panel thickness a load case was set up where half of the weight of the robot
(20 kg) is supported at one point in the middle of a 25 cm wide strip of ABS (Figure 86). It
should be noted that this is a very pessimistic scenario as the shell is supported by the
chassis along the most of the bottom of the robot.
W
t
L
Figure 86. Load on a shell panel modelled as a simply supported beam. L – width of the robot, t – shell
thickness, W – centrally applied load.
The stress induced in the material can be estimated using simple beam bending theory:
94
where
– stress induced in the sheet (
applied (
), – thickness of the sheet ( ),
), – second moment of area of the cross section (
– moment
). The maximum moment
arising due to a centrally applied load on a simply supported beam can be calculated by
where
– force applied ( ) and
– length of the unsupported section ( ). The second
moment of area for a rectangular cross section can be determined by
where
– breadth of the section ( ). Combining and rearranging the above equations an
expression for the stress can be obtained:
Assuming a load of 20 kg (200 N) over the width of the robot of 250 mm for a sheet
thickness of 2 mm and strip breadth of 250 mm the maximum stress induced in the material
is:
This is above the yield strength of ABS sheet (65 MPa (S&S 2013)). Increasing the sheet
thickness to 3 mm:
This results in a safety factor of 1.97 for a very unfavourable load case. Thus it can be
assumed that a 3 mm thick ABS sheet will be sufficient to withstand loads seen in service.
95
Appendix B: Power System support materials
B.1 Wiring Diagram
96
Appendix C: Sensors & devices supporting data
C.1 Capability analysis of existing sensors
Table 16. Capability analysis of existing sensors.
Capability
Devices/ Sensors
Front and back 2x M1054 IP cameras (Axis 
real-time view
Communications 2013)



Two-way
Communication
Axis M1054 IP camera – the 
front camera has 2-way audio 
Controlling arm 3x servomotors:

positions
 A-Max 26
 A-Max RE-36 and
 A-Max RE-30 (Maxon motor
ag 2013)
3x potentiometers:
 Vishay Model 357 (Vishay
2009)
Controlling
2x relative encoders:

flipper positions  Maxon HEDL-5540
Driving
Picking
objects

Maxon RE-50
up Off-the-shelf gripper powered 
by a HS-422 (Hitec 2013)
Heat detection
Photon 160 (FLIR Systems Co 
Ltd. 2012)
97
Strengths
Weaknesses
Low lag
 Occasional
dropouts
Clear front and
rear view of  Cameras are
surroundings
bulky
High
quality
image
Cameras have
built in LED
lights
Low lag
High
quality
audio
Potentiometers
measure
absolute
positions
–
suitable
for
inverse
kinematics
Flippers can be 
moved
in
incremental
steps
Full range of
mobility
and
speeds
Capable
of 
picking
up
bottles of water
and blocks with
eyelets
Capable
of 
detecting victim
heat
Encoders
relative
not
absolute
–
cannot
keep
track of flipper
positions
Gripper
attached
robot
not
to
Infrared
camera
not
functional on
robot
Futurlec
CO2
sensor 
(Futurlec 2013)
 Phidget board (Analogueto-digital-converter)
(Phidgets Inc. 2013)
Mapping/ point LiDAR module - Hokuyo URG- 
cloud data
04LX (Hokuyo Automatic Co.
2009)

CO2 sensor in 
working
condition
Xsens MTi-28A53G35 (Xsens 
Technologies B.V. 2013)
Capable
of 
providing
accurate
orientation data 
Breath
detection
Robot
Orientation

LiDAR
is 
functional
Capable
of
producing maps 
and point cloud
data


Head control – 2x
RX-64
servomotors 
pan/ tilt
(ROBOTIS 2010) using an RS485 bus
98
Pan and tilt of 
head
is
functional
CO2
sensor
incorrectly
powered
in
robot head –
not functional
LiDAR
adds
weight to the
head
LiDAR attached
to robot but
unpowered
No
mapping
function on the
client
Xsens
adds
weight to the
head
Xsens attached
to the robot
but unpowered
No orientation
data sent to
client
Occasional
dropouts
of
motors
–
power issue
Appendix D: Control system support materials
D.1 Inverse Kinematics – Flying the Head
The concept of flying the head is to move forwards, sideward and upwards based upon the
direction the head is facing.
In order to do this, a local cylindrical co-ordinate frame (r’-z’- ) is created on the joint by
which the head rotates. This co-ordinate frame is in the same orientation as the global coordinate frame (r-z- ).
Assuming that the arm begins is the positions
forward by a distance
and
, the change needed in both r(
and the user wants to move
) and z(
) can be calculated as
follows.
Firstly,a solution is found in the cylindrical co-ordinate frame.
z
z’
𝑑
𝜃𝑡
𝑝𝑧
r’
𝑝𝑟
𝑝𝑧
𝑑
𝑝𝑟
r
Figure 87 Robot arm in cylindrical coordinate frame
From this,
and
can be calculated in terms of the tilt angle
99
,
The cylindrical co-ordinate frame can now be transformed into a new Cartesian co-ordinate
frame (x’-y’-z’) using the pan angle
(where
and
are the current arm co-ordinates in
x and y respectively).
Figure 88 Robot arm top down view (x-y coordinate frame)
The Cartesian change in position can now be calculated in terms of the distance
angle
, and the tilt angle
The final position based upon the previous position can now be calculated,
100
, the pan
These are the absolute positions that are configured in the code and then sent to the arm.
This completes the process for flying the head forwards. For flying sideward and upwards,
the local co-ordinate frame on the end-effector is transformed using the appropriate
rotation matrices.
101
D.2 Controller layout
The main controller layout is shown in Figure 89. Several different modes were used in order
to fit all possible functions onto the one controller.
Figure 89. Main controller layout.
Drive Mode
The drive mode has been kept relatively the same as last year. This mode allows the user to
control each track separately, like a tank and is very effective when attempting to navigate
tight areas and step fields.
102
Figure 90. Drive mode controller layout (“tank” controls).
Inverse Kinematics Mode
The new inverse kinematics has been added as a separate arm control mode (Figure 91).
This mode allows the user to ‘fly’ the head and is intended for delicate control of the arm
such as pick and place operations or accurate positioning of the cameras for searching.
Figure 91. Inverse kinematics controller layout.
Search Mode
The aim of this mode was to replicate the controller layout familiar to most from shooting
video games (Figure 92). This allows the user to move the arm and tracks at the same time
and is useful for general driving of the robot and searching for victims in larger areas.
103
Figure 92. Search mode controller layout.
Joint-Control Mode
The ability to control each joint separately (Figure 93) was maintained this year as it has
proved very useful in the past, especially when attempting to manoeuvre the arm into
unusual and complicated positions.
Figure 93. Joint control mode layout
104
Appendix E: Other support materials
E.1 Remotec feedback
Peter Green
Programme and Future Development Manager
Northrop Grumman Corporation
Information Systems Europe
Remotec UK
29th April 2013 by email
Quote:
“The WMR system performed well for its first time out, there were concerns about its
reliability however railway lines were negotiated with little effort followed by some offroading and then stairs all within minutes of arrival.
The robot took some knocks and bumps and continuous vibration seen from track patter
throughout the day and kept going proving to be robust.
The team had obviously spent some time building up the new chassis which looked together,
and was easily accessible as it should be especially during the trialling period.
The team learnt that debris entering the track system will exert some forces on the chassis
that static stress modelling will not show. A bracing bar to maintain the track drive train
spacing may be needed as debris entering the track will locally bend the chassis rather than
stretch the rubber track.
As the day went on driver confidence grew and negotiation of terrain and use of arm
became second nature, again some valuable lessons were learnt here.
The Mimic ‘graphical model’ was very useful as the operator needed to know the
orientation of his robot to ensure stability on a number of occasions.
105
The environmental situation awareness information was also useful to fill in the picture for
the operator of distance and height which is often difficult to assess using 2d images from
cameras alone.
We would suggest possibly overlapping the mimic with this so that the robot is seen as part
of this overall sitrep rather than as a radar image, when the arm deploys the arm could be
seen to avoid the walls as an example.”
106
E.2 Recommendations for further work
Chassis

Design a method for keeping the tracks clear of debris

A bracing bar to maintain the track drive train spacing may be needed as debris
entering the track will locally bend the chassis rather than stretch the rubber track.

Further weatherproofing – start by sealing shafts and making shell airtight

Improve modulation of design – ensure each component is removable
independently of other parts
Power System

Fuses could be included as part of the housing instead of being located inside the
robot. This would improve protection against the housing connector being shortcircuited and make replacing fuses easier.

The inward-facing surfaces for mounting electronics to the battery case leave a very
small gap for wiring. Redesigning so that the electronics sit along one side of the
housing would make wiring easier.

The NFET disconnects the negative side of the battery, meaning the monitor’s
ground is no longer connected to the robot’s ground and so the communication line
between the two has to be isolated. A PFET solution would remove this
requirement.

Active balancing could be implemented to extend battery life.

The mbed is not designed for low-power operation. Replacing it with a smaller
microcontroller with a sleep mode would make the power board more efficient.

The I2C switch used to provide 4 ports was difficult to source, finding an alternative
or reducing the number of ports would solve this problem.
107

A number of wires and connectors have a much higher current capacity than
necessary. Using lower-spec components would reduce the size and weight taken up
by wiring.

Consider replacing the star-point ground with a second busbar to make rewiring
easier.

Try to have all connections to a system run through a single connector placed near
that system to make disconnecting it simple.

Investigate using SolidWorks Routing plug-in (Dassault Systemes 2012) to streamline
wiring harness design process

Redesign the motor controller boards to a smaller, standardised form factor. Keep
all connectors on one side of the board for easier wire management.
Sensors and Devices

Investigate longer range communication systems

Reduce weight from the arm by replacing heavy aluminium components with
lightweight alternatives

Source a smaller computer
Control System

Graceful communications failure – heartbeat or basic autonomy (return into radio
coverage)

Possible overlap of the 3D representation with the environmental situation
information (e.g. LiDAR feedback) so the robot is seen as part of the overall site

Investigate methods for autonomy – PieEye (08/09 report)

Blob detection for the IR camera – could overlay onto webcam images
108
E.3 List of Abbreviations
ABS - Acrylonitrile butadiene styrene
DoF – Degrees of Freedom
FDM – Fused Deposition Modelling
FEA – Finite Element Analysis
GUI – Graphical User Interface
IC – Integrated Circuit
IK – Inverse Kinematics
IMU – Inertial Measurement Unit
IP – Internet Protocol
IR – infrared
LED – light emitting diode
LiDAR - Light Detection and Ranging
LiPo – Lithium Polymer Battery
NFET - Negative Channel Field Effect Transistor
PFET - Positive Channel Field Effect Transistor
PID – Proportional, Integral and Derivative control
PS3 – PlayStation 3
PWM – Pulse Width Modulation
QR – Quick Response code
RLLC – Robot Low Lag Connection
SLAM – Simultaneous Localisation and Mapping
SWOT – Stregths, Weaknesses, Opportunities, Threats
USAR – Urban Search and Rescue
USB – Universal Serial Bus
VLC – Video LAN Client
WMR – Warwick Mobile Robotics
109